Waiting for rate cuts is how you miss the market

Rates went up. Why didn't the market break?

From 2022 the Federal Reserve raised its policy rate sharply and shrank its balance sheet at the same time. By the textbook, that is a hostile environment for equities, and especially for growth stocks, which are valued by discounting distant future profits back to the present.

The outcome was more complicated. US equities absorbed the tightening shock, and megacap technology moved to the center of a new investment cycle built around AI.

If you file that away as "stocks were strong despite high rates," you will make the same mistake in the next cycle. There is a better question than the level of rates. While the Fed was withdrawing money, where did the market's cash actually leave from, and where did it go?

The second hand on the money supply: the US Treasury

The US government keeps a huge operating account at the Fed called the Treasury General Account, or TGA. Tax receipts and the proceeds of debt issuance flow in. Government spending flows out.

When the TGA balance rises, cash moves out of the private financial system and into the government's account, which drains reserves. When the government spends TGA money, that cash returns to the private sector. The Fed sets the policy rate, but the timing of when cash leaves the financial system and when it comes back is also shaped by the Treasury's cash management.

Then there was the overnight reverse repo facility, the ON RRP. It is the window where eligible institutions such as money market funds park spare cash at the Fed overnight. In early 2023 roughly 2 trillion dollars was sitting there. As the Treasury then issued short-term bills on a large scale, money market cash drained out of the RRP and moved into T-bills, which paid more.

The important detail is that this money did not go straight into stocks. The primary destination of the falling RRP balance was mostly short-term Treasury paper. But because RRP cash absorbed that supply, the Treasury's borrowing did not suck bank reserves out all at once. That is one reason the liquidity squeeze the market felt during quantitative tightening was milder than the raw balance sheet numbers suggested.

The 2023 cushion is almost gone

This is the part that matters most right now. On 8 September 2026 the ON RRP balance stood at roughly 626 million dollars. It is effectively empty. The TGA, by contrast, averaged about 967.9 billion dollars in the week to 2 September.

So the argument that "there are still trillions in the RRP, so QT can keep being absorbed" no longer holds. The plumbing of liquidity in 2023 and the plumbing today are not the same.

That is not a prediction of a liquidity crisis. The point is simpler. From here, the effect of Treasury cash building and debt issuance on private bank reserves has to be watched more directly than before. Even an identical rise in the TGA hits the market differently depending on which investors the cash came from and what maturities the funding was raised in.

Washington would rather grow the denominator than pay down the debt

The core problem in US public finance is not the absolute size of the debt but the debt relative to the size of the economy. The CBO projects federal debt held by the public will reach about 101% of GDP by the end of 2026. The deficit runs at roughly 1.9 trillion dollars, or 5.8% of GDP.

Cutting the debt quickly would require tax increases or large spending cuts. Neither is politically easy. Inflating away the real value of the debt is another route, but it pushes up long-term rates and the cost of living.

That leaves the least painful option: growing the denominator, meaning productivity and nominal GDP. This is why the United States treats AI as national growth infrastructure. If AI does more than move technology share prices, and actually improves labour productivity and corporate cost structures, then the same debt can be carried by a larger economy.

But that is still a hypothesis, not a result. Claiming AI has already lifted economy-wide productivity ignores how much real friction remains in data integration, security, redesigning workflows, and power and network capacity.

AI spending looks less like an app subscription and more like building a factory

The defining feature of this AI cycle is that it is a software boom whose money pours into physical assets. GPUs, CPUs, HBM, servers, network gear, data center buildings, substations and generation capacity are all needed at once.

Microsoft recorded 41 billion dollars of capital expenditure in the fourth quarter of its 2026 fiscal year alone, and said about two thirds of it went into relatively short-lived assets such as GPUs and CPUs. The company guided to roughly 190 billion dollars of capex for the 2026 calendar year as a whole.

Those numbers mean AI is no longer a research line item at a handful of software companies. It is a large industrial equipment cycle. And the real contest starts now. That equipment has to prove how much it lifts revenue in Azure, advertising, search, Copilot and enterprise AI, and how much it cuts labour costs and working hours.

Having built a data center is not productivity. The compute inside the building becomes productivity only when it earns customers money or removes their costs.

Watch the reason rates moved, not the rate

Long-term yields cause the same error. Judging whether stocks live or die by a single number does not work. A 10-year yield of 4.8% tells you very little on its own. The reason matters.

If it reflects better expectations for growth and productivity, and therefore a stronger nominal growth outlook, high rates can coexist with strong earnings growth. If it is rising because of fiscal deficits, the burden of Treasury supply and an inflation risk premium, that is far worse for equity valuations and for corporate funding costs.

Which is why the two-line formula, rates up equals stocks down and rate cuts equal stocks up, is nearly useless for investing. Rates are not only a cause. They are also the result of several forces working inside the economy.

What to Watch

First, the TGA balance. If the government accumulates cash quickly, it can work in the direction of draining private reserves. Read it alongside the maturity structure of debt issuance rather than as a single number.

Second, bank reserves and money market rates. With the RRP cushion all but gone, the question is whether falling reserves show up as stress in short-term rates such as SOFR.

Third, the reason the 10-year yield is rising. Separate growth expectations from an inflation and fiscal premium.

Fourth, hyperscaler capex against AI revenue growth. Check whether Microsoft, Alphabet, Amazon and Meta are growing AI-related revenue and cash flow faster than they are growing investment.

Fifth, data center power and financing structures. The more you see grid connection delays, long-term PPAs, project finance and equipment-backed lending, the greater the chance technology risk turns into financial risk.

Insight Times Editorial Desk